Re: [patch] New thread test to exercise Daniel's Patch

[Daniel Jacobowitz <drow at false dot org>]

On Mon, Mar 29, 2004 at 04:59:30PM -0500, Jeff Johnston wrote:
> Daniel Jacobowitz wrote:
> >On Mon, Mar 29, 2004 at 04:04:00PM -0500, Jeff Johnston wrote:
> >
> >>Ok, I had the thread checks in the Continue test.  If I change it to the 
> >>way above, now I get 3 "Process no longer exists "messages but the test 
> >>completes. :(
> >>
> >>I'm really starting to hate these test macros.  What is the reasoning 
> >>behind avoiding send_gdb / gdb_expect?
> >
> >
> >They don't automatically handle things like internal errors,
> >disconnects, et cetera.  Want to post your current version and I'll
> >give it a try?
> >

>   -re "Program received signal SIGINT.*$gdb_prompt $"
>     {
>       pass "$message"
>     }

Nothing automatically sets $message.  The ERROR: you saw is DejaGNU's
generic failure message for things like syntax errors in expect blocks;
I sent a patch to the dejagnu list a month or so ago to print more
helpful information.  Try the attached script instead.

Oddly, running a fixed manythreads.exp with an unpatched GDB, I get a
SIGSEGV in pthread_join.  It shows up as a FAIL (yay).  The patched GDB
shows up as nine PASSes (yay).  Re-running it a number of times, the
SIGSEGV came and went intermittently.

Running the test with LinuxThreads an internal error (lp->status == 0
assertion failed) came and went also.

I guess that makes it a good test.... now someone will have to _fix_
those.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Attachment: manythreads.exp
Description: Text document